What’s that bit about the thesis’ deadline? Is Matz working on a PhD?

Any ideas about features for a new GC are for Ruby? Any fix for continuations + weakrefs?

babie: thanks one more time, your translation effort is unvaluable :)

Anyway, I was thinking if, while in the process of writing a new GC, could it be possible to have some kind of “pluggable gc” framework.

For example, since many projects use the Bohem’s GC (i.e. GCJ relies on it) it could make sense to make use of it when it is available (but maybe not)

Boehm’s GC, although it optionally supports types, can almost certainly be bettered by a GC written specifically to handle a strongly-typed language.

Boehm’s GC is a hack, like any conservative GC. A very useful one, but it is a hack.

A precise GC is much much better if your language supports it—the only reason for Boehm’s conservative approach is that C can’t support precise GC without writing manual marking routines for everything.

StUdLy CaPS are a hack. A very silly one, but they are a hack.

”Boehm’s GC is a hack, like any conservative GC”

We prefer to call it a heuristic around these here parts…

I’m not much into GCs (I’m not much into anything, except in my body and not necessarily there) but I don’t understand how strong typing makes bohem’s gc bad. It is used in quite a lot of strongly typed systems (SmartEiffel can use it, mono and gcj use it, Goo too). Maybe you mean writing it in a strongly typed language, but then, I think much more has to come before we see it in ruby.

Anyway, I’m sure there can be better GCs, this why I think a pluggable memory manager would be ubercool :)

Sorry, I didn’t read these comments till today. I will reply these and translate the related article after I go home. Please wait.